I am usıng Couenne solver to solve my MINLP problem. I found optimal solution for my small size problems. However I could only find integer solutions for my medium size problems in less than 2 seconds. I guess the reason of that is the NLP heuristic does not run long enough. Because I got this line in the log:

NLP Heuristic: time limit reached.

I tried several options to increase the run time but I could not run the NLP heuristic longer? Which options should I change?

According to the documentation you can use reslim (in seconds). “GAMS/Couenne understands currently the following GAMS parameters: reslim (time limit), nodlim (node limit), cutoff, optca (absolute gap tolerance), and optcr (relative gap tolerance). Further, the option threads can be used to control the number of threads used in the linear algebra routines of IPOPT.” This means that you don’t have to write an option file, but can set the option like this: mymodel.reslim = 100000; or option reslim = 100000;

Thank you for your reply. I tried all of the parameters that you sent and I got the following results and questions.

Currently my model status has 8 integer solutions and the log file says "NLP heuristic: time limit reached ". When I changed the parameter of “warm_start_init_point” as “YES” then the log file says “no solution”.

2)After I set the parameters of optca=0 and optcr=0.1 then, I still have 8 integer solutions. If I set the parameters of optca=0 and optcr=0 then, I have 7 feasible solutions. Why do I loose the integer solutions and get feasible solutions? Should not I get optimal with the settings of optca=0 and optcr=0?

Also is it possible to say that the integer solutions with the settings of optca=0 and optcr=0.1 are close to the optimal with 10%?

Why do I get integer solutions with Couenne solver, which should guarantee the optimality for non-convex MINLP?